April 2020 - webkit-dev - lists.webkit.org

Request for feedback on CompressionStream and DecompressionStream
by Adam Rice 02 May '22

02 May '22

I am trying to gauge feedback on compression streams with a view to shipping them in Chromium. Very briefly, they are a way to do gzip and gunzip in the browser. Less briefly, the explainer https://github.com/WICG/compression/blob/master/explainer.md goes into some detail of the how and why. The specification https://wicg.github.io/compression/ gives verbose detail of how. You may also find the W3C TAG review https://github.com/w3ctag/design-reviews/issues/410 interesting. Any and all feedback is welcome! FAQ: Q. Will this support brotli, zstd, etc.? A. Not yet. It's on the radar. Thanks, Adam Rice, Chromium

3 3

Feedback on the performance.measureMemory API proposal
by Ulan Degenbaev 14 Dec '20

14 Dec '20

Hello webkit-dev, I'd like to solicit feedback as well as an official position from Webkit on our proposal for the new memory measurement API: https://github.com/WICG/performance-measure-memory The proposal generalizes the previous performance-memory proposal [1] that was abandoned [2] due to information leak of cross-origin resources. The new API addresses the information leak and process model issues raised in the previous feedback [3]. What is different in the new API? - It relies on COOP+COEP [4] for preventing cross-origin information leaks (see "Security Considerations"). Besides security, COOP+COEP also helps implementability because the browser needs to isolate the web page from other web pages. - It does not assume any particular process model and is defined using the standard concepts such as JS agent cluster and browsing context group. - It has an extensible and generic interface for breaking down memory usage by type (JS, DOM, GPU) and owner. Implementers can choose the level of detail they are willing to provide based on their performance trade-offs (see "Performance Considerations"). Thank you, Ulan. [1]: The previous proposal: https://github.com/WICG/performance-memory/blob/master/explainer.md [2]: Information leak affecting the previous proposal: https://github.com/mozilla/standards-positions/issues/85#issuecomment-42638… [3]: Feedback from Apple on the previous proposal: https://github.com/WICG/performance-memory/issues/14 [4]: COOP and COEP explained: https://docs.google.com/document/d/1zDlfvfTJ_9e8Jdc8ehuV4zMEu9ySMCiTGMS9y0G… Other Relevant Links Explainer: https://github.com/WICG/performance-measure-memory Open issues: https://github.com/WICG/performance-measure-memory/issues WebPerf WG F2F June 2019 presentation (for older iteration of the API): https://docs.google.com/document/d/1uQ7pXwuBv-1jitYou7TALJxV0tllXLxTyEjA2n1… TAG design review (for older iteration of the API): https://github.com/w3ctag/design-reviews/issues/386 Issue on Mozilla standards-positions: https://github.com/mozilla/standards-positions/issues/281

1 1

Implementing MathML stylistic attributes in WebKit
by Frédéric Wang 28 Sep '20

28 Sep '20

Hello, Currently MathML attributes mathvariant, displaystyle and scriptlevel [1] [2] are implemented in WebKit using custom "style resolution" and "one-glyph rendering" inside the MathML layout code [3] [4] [5]. These features involve text rendering and interaction with CSS font-size, so it is difficult to implement them properly and completely that way. There are known bugs and missing features right now (e.g. mathvariant transform only applies to one-character strings, automatic displaystyle/scriptlevel does not work with fractions, etc) Several years ago, Gecko used to do something similar but that was causing a lot of problems (dynamic update, assertion failures...). At the end, this is now implemented in Gecko more reliably by mapping the attributes to internal CSS properties ; [6] is based on that. When I tried to do something like this in WebKit three years ago, I was only able to rely on proprietary -webkit-* extensions exposed to users [7]. So my questions are: (1) What is WebKit's mechanism to implement such stylistic attributes? (2) Is it possible to implement internal (i.e. not web-exposed to users) CSS properties/values in WebKit? (3) Is it ok to add more -webkit-* properties/values or should these properties be standardized at the CSS WG instead? I understand answers to these general questions from other browser project is either use internal properties or standardize them. Thanks, Frédéric PS: For a concrete example, see [8]. Automatic mathvariant is used for italic variables while font-size in fractions is calculated according to displaystyle / scriptlevel rules. The latter is incorrect (check the attached screenshot) while selection highlight is broken for the former. [1] https://mathml-refresh.github.io/mathml-core/#the-mathvariant-attribute [2] https://mathml-refresh.github.io/mathml-core/#the-displaystyle-and-scriptle… [3] https://trac.webkit.org/browser/webkit/trunk/Source/WebCore/css/mathml.css [4] https://trac.webkit.org/browser/webkit/trunk/Source/WebCore/rendering/mathm… [5] https://trac.webkit.org/browser/webkit/trunk/Source/WebCore/rendering/mathm… [6] https://mathml-refresh.github.io/mathml-core/#css-extensions-for-math-layout [7] https://bugs.webkit.org/show_bug.cgi?id=133845#c5 [8] https://bug-201242-attachments.webkit.org/attachment.cgi?id=377730 -- Frédéric Wang

2 4

PSA: Bit fields won't be packed on Windows if you mix types
by Ryosuke Niwa 04 Sep '20

04 Sep '20

Hi, Unlike gcc and clang, MSVC pads each consecutive member variables of the same type in bitfields. e.g. if you have: struct AB { unsigned m_1 : 31; bool m_2 : 1; } then *MSVC pads m_1 and allocates sizeof(unsigned) * 2 for AB* whereas gcc and clang only allocate sizeof(unsigned) * 1 for AB. This is *not a compiler bug*. It's a spec. compliant behavior, and may in fact have a better run-time performance in some situations. However, for core objects like RenderObject and InlineBox, allocating extra 4-8 bytes per each object is prohibitory expensive. In such cases, please *use the same POD type for all bitfield member variables*. (Storage class classifiers and variable qualifiers seem to have no effect on how variables are packed; e.g. mutable, const, etc...). For example, MSVC will allocate sizeof(unsigned) * 1 for AB if we rewrite the above code as: struct AB { unsigned m_1 : 31; unsigned m_2 : 1; } When you're making this change, *be sure to audit all code that assigns a non-boolean value to m_2* because implicit type coercion into boolean is no longer in effect. For example, AB ab; ab.m_2 = 2; puts(ab.m_2 ? "true" : "false"); will print "true" before the change and will print "false" after the change. An easy way to ensure you've audited all such code is to add getters and setters for all bitfield member variables or wrap them in a special structure or class as done in http://trac.webkit.org/changeset/103353 and http://trac.webkit.org/changeset/104254. Best regards, Ryosuke Niwa Software Engineer Google Inc.

4 8

Feedback on Blink's text fragment directive proposal
by David Bokan 31 Aug '20

31 Aug '20

Hello webkit-dev, I'd like to solicit feedback as well as an official position from Webkit on our proposal for the text fragment directive: https://github.com/WICG/ScrollToTextFragment. In summary: this is a feature that allows authors and users to craft URLs to pages and specify a snippet of text on the page as a subresource (visually highlighting it and scrolling it into view). Analogous to element-id based fragment anchors but for text. You can try this out today in Chrome Beta by enabling chrome://flags/#enable-text-fragment-anchor. Here's an example link: https://en.wikipedia.org/wiki/Uniform_Resource_Identifier#:~:text=An%20opti… Relevant Links: <https://github.com/WICG/ScrollToTextFragment/blob/master/README.md> Explainer <https://github.com/WICG/ScrollToTextFragment/blob/master/README.md> Spec <https://wicg.github.io/ScrollToTextFragment/> TAG Review <https://github.com/w3ctag/design-reviews/issues/392> (Currently Suspended) Blink Intent Thread <https://groups.google.com/a/chromium.org/forum/#!msg/blink-dev/zlLSxQ9BA8Y/…> Issue on Mozilla standards-positions <https://github.com/mozilla/standards-positions/issues/194> We've been using the GitHub repo for issue tracking but happy to take feedback (official or otherwise) in any form. Thank you! David

4 4

EWS now parses error logs in case of build failure
by Aakash Jain 11 Aug '20

11 Aug '20

Hi Everyone, I am happy to announce another EWS feature. From now on, in case of build failure, EWS will parse the errors and display them in a separate 'errors' log. You wouldn't have to search through thousands of lines of logs to find the error message. For example, in https://ews-build.webkit.org/#/builders/16/builds/6054, in step #7 WebKit failed to compile. Complete logs (stdio) are 38,000+ lines, and the error is not at the end of the logs. Normally, it requires some searching through the logs to find the relevant errors. But now, there is another 'errors' log, which contains just the relevant 11 lines (containing error and few related lines to provide additional context). Hopefully this would save some time and efforts previously spent on searching through the large logs. Note that this information is not displayed in status-bubble tool-tip, since this might be lot of text to display in the tooltip. My further plan is to make this information more readily available, by adding it to a custom designed page which will open on clicking the status bubble https://webkit.org/b/197522 Please let me know if you notice any issues or have any feedback. Thanks Aakash Reference: https://webkit.org/b/203418

4 6

PSA: Changes in the recommended GTK/WPE developer workflow
by Philippe Normand 07 May '20

07 May '20

Hi, Until now the most recommended way to work on the GTK and WPE ports was to use a custom JHBuild[0] moduleset ([1][2]) for build/runtime dependencies. This setup was deployed almost 10 years ago [3] and while it is an opt-in approach, it is highly recommended, especially if you want to run layout tests locally. If you ever executed Tools/Scripts/update-webkitgtk-libs or Tools/Scripts/update-webkitwpe-libs it means you currently have a DependenciesGTK and/or DependenciesWPE folder in WebKitBuild/ and that your workflow is JHBuild-based. JHBuild has various shortcomings, even though it served us well all these years: - if the moduleset is modified, the bots remove the whole Dependencies{GTK,WPE} folders and rebuild everything from scratch - no cross-compilation support - no clear separation between the host system and the JHBuild-based sysroot - poor system requirements management (See the install-dependencies scripts in Tools/{wpe,gtk}) - time lost rebuilding the dependencies (can take up to an hour on my build machine) So in order to improve our lives a bit, we decided to try a new workflow, this time based on Flatpak[4]. Instead of having every developer locally build the dependencies, the built sysroot will be packaged in a Flatpak SDK/Runtime and downloaded to the developer machine. Currently the SDK is built for X86_64 but additional architectures can be supported (ARMv7, Aarch64 at least). The advantages of this new approach: - cross-compilation can easily be achieved - clear separation between the host system and the SDK sysroot - the SDK runs in a sandbox - unified toolchain for WebKit build (currently both GCC 9.2.0 and clang 8.0.1) - no time lost rebuilding dependencies :) - consistent layout tests coverage across different hosts - separate build directories for each port (example: WebKitBuild/GTK/Release mounted as /app/WebKit/WebKitBuild/Release in the SDK sandbox) As a bonus, this setup allows for: - integration with sccache, for faster WebKit builds - builtin IceCC support without additional manual steps (you still need a scheduler running on the host system though) The only disadvantage of this new approach is that hacking on dependencies is currently not trivial to accomplish. We plan to enable this most likely via the Flapjack[5] tool. Once the patch from bug#205658 [6] lands, this workflow will be available. How do I use it? Easy: 1. Make sure your flatpak version is recent enough, 1.4.4 is the minimum version we support. Backports for Debian stable are available and for Ubuntu LTS, a PPA is available. See the Flathub instructions (though you don't need to manually add the flathub remote) [7]. 2. Run update-webkitgtk-libs or update-webkitwpe-libs as usual. This will download and install the SDK in WebKitBuild/UserFlatpak/ 3. Run build-webkit as usual The SDK will be used when running the tests (API, layout, webkitpy, etc) and the MiniBrowser. For the time being we keep the JHBuild workflow in the SVN, although if everything goes well, it will be removed in the coming months. Hence we encourage all GTK/WPE developers to try this new workflow! If you still want to keep using JHBuild, make sure to set the WEBKIT_JHBUILD=1 environment variable in your shell. But keep in mind that we intend to remove support for JHBuild if Flatpak works as expected (so please try to test it, and report any issue with it). The GTK/WPE buildbots and EWS are currently running with JHBuild and will be switched one-by-one to Flatpak soon. The SDK sources are currently hosted on Github[8], but we might move them to WebKit's SVN. Philippe [0] https://developer.gnome.org/jhbuild/stable/ [1] https://trac.webkit.org/browser/trunk/Tools/gtk/jhbuild.modules [2] https://trac.webkit.org/browser/trunk/Tools/wpe/jhbuild.modules [3] https://bugs.webkit.org/show_bug.cgi?id=73425 [4] https://flatpak.org/ [5] https://github.com/endlessm/flapjack [6] https://bugs.webkit.org/show_bug.cgi?id=205658 [7] https://flatpak.org/setup/ [8] https://github.com/igalia/webkit-flatpak-sdk

1 2

Position on emerging standard: WebCodecs
by Dan Sanders 06 May '20

06 May '20

Hello, I'm reaching out to see if WebKit would like to weigh in on the WebCodecs WICG proposal: https://discourse.wicg.io/t/webcodecs-proposal/3662. The WebCodecs API enables web developers to instantiate codecs (audio/video encoders/decoders) and use them to process individual frames. There is a related proposal for image decoders; it enables access to individual animation frames: https://discourse.wicg.io/t/proposal-imagedecoder-api-extension-for-webcode…. An implementation of these APIs is being developed in Chromium. Thank you, Dan Sanders

2 3

Exposing FeaturePolicy to the web
by Sergio Villar Senin 29 Apr '20

29 Apr '20

Hi, WebKit currently implements the feature policy mechanism[1] to allow developers/sites to enabled/disable certain features and/or browser APIs like geolocation or access to camera among others. While implementing the WebXR API I realized that we are not exposing[2] the values of feature policies to the web. I was wondering whether that was made on purpouse (fingerprinting?, privacy?...) or it is just because the use cases were not considered relevant enough to implement it. I haven't found any bug tracking that. BR [1] https://w3c.github.io/webappsec-feature-policy [2] https://w3c.github.io/webappsec-feature-policy/#the-policy-object

1 0

webkitgtk and bubblewrap help needed
by Jack Hill 28 Apr '20

28 Apr '20

Hi WebKit, I'm working on an issue [0] in GNU Guix's WebKitGTK package. In the issue, I've identified what I believe to be a problem at the intersection of Bubblewrap, WebKitGTK, and Guix. I believe that the source of the problem is that with Guix, files that are needed in the sandbox are located in unexpected places. Everything manages by Guix, such as packages and configuration files, gets its own path under /gnu/store (answering why it is this way is a longer discussion which I'll leave to the manual [1] for now, but I find it compelling). Some files and directories under /run and /etc are symlinks to their canonical location under /gnu/store. I believe that this arrangement will require extra paths to be bound into Bubblwrap's new mount namespace. However, it isn't clear to me what they are. I have tried patching BubblewrapLauncher.cpp, but it still seems to get tripped up with the pulse/client.conf symlink. I have not been able to reproduce the problem with simple interactive invocations of bwrap. I'm interested in hearing how to troubleshoot problems like: bwrap: Can't create file at /etc/pulse/client.conf: No such file or directory bwrap: Can't mkdir parents for /run/current-system/profile/lib/gstreamer-1.0: No such file or directory I'm also interested in hearing how to best integrate WebKitGTK with Guix. I suppose the ideal case is for WebKitGTK to detect situations like Guix's and call bwrap with the correct arguments, and to fall back to carrying a local patch in Guix if a general solution is not appropriate. [0] https://debbugs.gnu.org/cgi/bugreport.cgi?bug=40837 [1] https://guix.gnu.org/manual/en/html_node/Managing-Software-the-Guix-Way.html Thanks, Jack

2 3